In the acquisition of geometric data, accuracy can sometimes be sacrificed for convenience and productivity. With this situation in mind, a simple system of automatic geometric data acquisition is developed. The idea is similar to the convensional method of silhouettes. Instead of using the silhouette, however, video images are used. The object to be measured is required to have a certain convexity property.
The video image taken at different angle is processed to obtain the boundary of the object. An image in the camera can be considered as the cross-section of a unbounded cone with its vertex at the focal point of the lense of camera at a specified location. For each elevation an approximate polyhedral shape which encompasses the object is obtained by intersecting these cones emanating from a set of different vertex locations. The largest convex smooth shapes enclosed by these polygons represent a cross-section of the object. A series of such cross-sections for a number of elevations are then used to construct the three-dimensional shape of the object.
The range of errors due to limitation in the hardware of the image capture system and due to the specific way of geometric construction are analyzed under specific assumptions. The developed system is tested for various objects such as pentagonal cylinder, twisted rectangular vase, foot and leg of mannequin and torso of a venus statue. The performance has been satisfactory as originally aimed.